KnowMe-Bench: Benchmarking Person Understanding for Lifelong Digital Companions

TL;DR

KnowMe-Bench is a new benchmark using autobiographical narratives to evaluate person understanding in AI, emphasizing stable motivations and decision principles beyond simple retrieval accuracy.

Contribution

It introduces a novel long-form autobiographical narrative benchmark with evidence-linked questions for comprehensive person understanding evaluation.

Findings

Retrieval-augmented systems improve factual recall.

Errors remain in temporally grounded explanations.

Memory mechanisms beyond retrieval are needed.

Abstract

Existing long-horizon memory benchmarks mostly use multi-turn dialogues or synthetic user histories, which makes retrieval performance an imperfect proxy for person understanding. We present \BenchName, a publicly releasable benchmark built from long-form autobiographical narratives, where actions, context, and inner thoughts provide dense evidence for inferring stable motivations and decision principles. \BenchName~reconstructs each narrative into a flashback-aware, time-anchored stream and evaluates models with evidence-linked questions spanning factual recall, subjective state attribution, and principle-level reasoning. Across diverse narrative sources, retrieval-augmented systems mainly improve factual accuracy, while errors persist on temporally grounded explanations and higher-level inferences, highlighting the need for memory mechanisms beyond retrieval. Our data is in \href{KnowMeBench}{https://github.com/QuantaAlpha/KnowMeBench}.